Kinetic study in an automatic continuous-flow photochemical platform with machine learning

In this work, we first solved the partial differential equation of the one-dimensional axial diffusion model in an open-source platform, that is, FEniCS, to explore the influence of the axial dispersion on the reaction yield-to-time profile. Then, we built an automatic platform, which included a photomicroreactor, a continuously controlled pump, a high-power UV-LED light source, an in-line visible-light absorbance analytical unit, and a Raspberry Pi based controlling unit. Moreover, steady-state feeding and sampling functions could be realized in this continuous-flow photochemical platform. The homogeneous photolysis of methylene blue and the photo-Favorskii rearrangement synthesis of ibuprofen as the model reactions were used to validate the robustness of this automatic platform with unsteady-state and steady-state operations, through which kinetic study was carried out using genetic algorithm based symbolic regression, leading to deep understanding on reaction mechanisms and benefits for process optimization.